Spaces:
Running
Running
File size: 50,311 Bytes
a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 02817d5 a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f 02817d5 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 f325822 a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f 02817d5 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 c841e1f a5b1635 fb1c81d 3eebcac fb1c81d a5b1635 fb1c81d c841e1f fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d c841e1f fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d c841e1f fb1c81d 58b94ba c841e1f fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d c841e1f fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d c841e1f a5b1635 fb1c81d c841e1f fb1c81d a5b1635 fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d a5b1635 fb1c81d c841e1f fb1c81d f325822 fb1c81d f325822 fb1c81d f325822 fb1c81d c841e1f fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d a5b1635 fb1c81d a5b1635 c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d c841e1f fb1c81d a5b1635 fb1c81d a5b1635 afff948 80a1295 bd5bd8e 80a1295 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 |
import os
import json
import torch
import logging
import traceback
from typing import Dict, List, Optional, Tuple
import time
from datetime import datetime
import threading
from collections import defaultdict
import gradio as gr
import numpy as np
import librosa
import soundfile as sf
from pydub import AudioSegment
from audio_separator.separator import Separator
from audio_separator.separator import architectures
class AudioSeparatorD:
def __init__(self):
self.separator = None
self.available_models = {}
self.current_model = None
self.processing_history = []
self.model_performance_cache = {}
self.model_recommendations = {}
self.setup_logging()
self.model_lock = threading.Lock()
def setup_logging(self):
"""Setup logging for the application"""
logging.basicConfig(level=logging.INFO)
self.logger = logging.getLogger(__name__)
def get_system_info(self):
"""Get system information for hardware acceleration"""
info = {
"pytorch_version": torch.__version__,
"cuda_available": torch.cuda.is_available(),
"cuda_version": torch.version.cuda if torch.cuda.is_available() else "N/A",
"mps_available": hasattr(torch.backends, "mps") and torch.backends.mps.is_available(),
"device": "cuda" if torch.cuda.is_available() else ("mps" if hasattr(torch.backends, "mps") and torch.backends.mps.is_available() else "cpu"),
}
# Only add memory info if CUDA is available
if torch.cuda.is_available():
info["memory_total"] = torch.cuda.get_device_properties(0).total_memory
info["memory_allocated"] = torch.cuda.memory_allocated()
else:
info["memory_total"] = 0
info["memory_allocated"] = 0
return info
def analyze_audio_characteristics(self, audio_file: str) -> Dict:
"""Analyze audio file characteristics for smart model selection"""
try:
# Load audio for analysis
y, sr = librosa.load(audio_file, sr=None)
duration = len(y) / sr
# Analyze spectral characteristics
spectral_centroids = librosa.feature.spectral_centroid(y=y, sr=sr)[0]
spectral_rolloff = librosa.feature.spectral_rolloff(y=y, sr=sr)[0]
zero_crossing_rate = librosa.feature.zero_crossing_rate(y)[0]
# Analyze tempo and rhythm
tempo, _ = librosa.beat.beat_track(y=y, sr=sr)
# Analyze dynamic range
rms = librosa.feature.rms(y=y)[0]
dynamic_range = np.std(rms)
# Determine audio characteristics
characteristics = {
"duration": duration,
"sample_rate": sr,
"tempo": float(tempo),
"avg_spectral_centroid": float(np.mean(spectral_centroids)),
"avg_spectral_rolloff": float(np.mean(spectral_rolloff)),
"avg_zero_crossing_rate": float(np.mean(zero_crossing_rate)),
"dynamic_range": float(dynamic_range),
"audio_type": self._classify_audio_type(
np.mean(spectral_centroids),
float(tempo),
dynamic_range
)
}
return characteristics
except Exception as e:
self.logger.error(f"Error analyzing audio: {str(e)}")
return {"audio_type": "unknown", "error": str(e)}
def _classify_audio_type(self, spectral_centroid: float, tempo: float, dynamic_range: float) -> str:
"""Classify audio type based on spectral and temporal features"""
if spectral_centroid < 1000:
return "bass_heavy"
elif spectral_centroid > 4000:
return "bright_crisp"
elif tempo > 120:
return "upbeat"
elif dynamic_range > 0.1:
return "dynamic"
else:
return "balanced"
def get_available_models(self):
"""Get list of available models with enhanced information"""
try:
with self.model_lock:
if self.separator is None:
self.separator = Separator(info_only=True)
models = self.separator.list_supported_model_files()
simplified_models = self.separator.get_simplified_model_list()
# Enhance model information
enhanced_models = {}
for model_name, model_info in simplified_models.items():
# Parse model filename for better names
friendly_name = self._generate_friendly_name(model_name, model_info)
# Determine best use cases
use_cases = self._determine_use_cases(model_name, model_info)
# Estimate performance characteristics
perf_chars = self._estimate_performance(model_name)
enhanced_models[model_name] = {
**model_info,
"friendly_name": friendly_name,
"use_cases": use_cases,
"performance_characteristics": perf_chars,
"architecture_type": self._get_architecture_type(model_name),
"recommended_for": self._get_recommendations(model_name, model_info)
}
return enhanced_models
except Exception as e:
self.logger.error(f"Error getting available models: {str(e)}")
return {}
def _generate_friendly_name(self, model_name: str, model_info: Dict) -> str:
"""Generate user-friendly model names"""
# Remove common prefixes and suffixes
clean_name = model_name.replace('model_', '').replace('.ckpt', '').replace('.yaml', '')
# Handle specific known models
if 'roformer' in model_name.lower():
return f"π΅ Roformer {clean_name.split('_')[-1] if '_' in clean_name else ''}".strip()
elif 'demucs' in model_name.lower():
return f"π₯ Demucs {clean_name.replace('htdemucs', '').replace('_', ' ')}".strip()
elif 'mdx' in model_name.lower():
return f"π€ MDX-Net {clean_name[-3:] if clean_name[-3:].isdigit() else ''}".strip()
else:
# Capitalize words
words = clean_name.replace('_', ' ').split()
return ' '.join(word.capitalize() for word in words)
def _determine_use_cases(self, model_name: str, model_info: Dict) -> List[str]:
"""Determine what this model is best for"""
use_cases = []
# Check output stems
if 'vocals' in str(model_info).lower():
use_cases.append("π€ Vocal Isolation")
if 'drums' in str(model_info).lower():
use_cases.append("π₯ Drum Separation")
if 'bass' in str(model_info).lower():
use_cases.append("πΈ Bass Extraction")
if 'instrumental' in str(model_info).lower():
use_cases.append("πΉ Instrumental")
if 'guitar' in str(model_info).lower() or 'piano' in str(model_info).lower():
use_cases.append("πΈ Specific Instruments")
# Architecture-based use cases
if 'roformer' in model_name.lower():
use_cases.append("β‘ High Quality")
elif 'demucs' in model_name.lower():
use_cases.append("ποΈ Multi-stem")
elif 'mdx' in model_name.lower():
use_cases.append("π΅ Fast Processing")
return use_cases[:3] # Limit to top 3
def _estimate_performance(self, model_name: str) -> Dict:
"""Estimate performance characteristics"""
perf = {
"speed_rating": "medium",
"quality_rating": "medium",
"memory_usage": "medium"
}
if 'roformer' in model_name.lower():
perf.update({"speed_rating": "slow", "quality_rating": "high", "memory_usage": "high"})
elif 'demucs' in model_name.lower():
perf.update({"speed_rating": "slow", "quality_rating": "high", "memory_usage": "high"})
elif 'mdx' in model_name.lower():
perf.update({"speed_rating": "fast", "quality_rating": "medium", "memory_usage": "low"})
return perf
def _get_architecture_type(self, model_name: str) -> str:
"""Extract architecture type from model name"""
if 'roformer' in model_name.lower():
return "π΅ Roformer (MDXC)"
elif 'demucs' in model_name.lower():
return "π₯ Demucs"
elif 'mdx' in model_name.lower():
return "π€ MDX-Net"
elif 'vr' in model_name.lower():
return "ποΈ VR Arch"
else:
return "π§ Unknown"
def _get_recommendations(self, model_name: str, model_info: Dict) -> Dict:
"""Get specific recommendations for model usage"""
recommendations = {
"best_for": "General use",
"avoid_for": "None",
"tips": []
}
if 'roformer' in model_name.lower():
recommendations.update({
"best_for": "High-quality vocal isolation",
"avoid_for": "Real-time processing",
"tips": ["Best results with longer audio files", "Higher memory usage", "Excellent for final mastering"]
})
elif 'demucs' in model_name.lower():
recommendations.update({
"best_for": "Multi-stem separation (drums, bass, vocals)",
"avoid_for": "Simple vocal/instrumental separation",
"tips": ["Creates multiple output files", "Good for music production", "Slower but comprehensive"]
})
elif 'mdx' in model_name.lower():
recommendations.update({
"best_for": "Fast vocal isolation",
"avoid_for": "Multi-instrument separation",
"tips": ["Quick processing", "Good for demos", "Lower memory requirements"]
})
return recommendations
def auto_select_model(self, audio_characteristics: Dict, desired_stems: List[str],
priority: str = "quality") -> Optional[str]:
"""Automatically select the best model based on audio characteristics and requirements"""
try:
models = self.get_available_models()
if not models:
return None
# Score models based on criteria
model_scores = {}
for model_name, model_info in models.items():
score = 0
# Base score from performance characteristics
perf_chars = model_info.get('performance_characteristics', {})
if priority == "quality":
if perf_chars.get('quality_rating') == 'high':
score += 10
elif perf_chars.get('quality_rating') == 'medium':
score += 5
elif priority == "speed":
if perf_chars.get('speed_rating') == 'fast':
score += 10
elif perf_chars.get('speed_rating') == 'medium':
score += 5
# Audio type matching
audio_type = audio_characteristics.get('audio_type', 'balanced')
use_cases = model_info.get('use_cases', [])
if audio_type == 'bass_heavy' and 'πΈ Bass Extraction' in use_cases:
score += 8
elif audio_type == 'bright_crisp' and 'π€ Vocal Isolation' in use_cases:
score += 8
elif audio_type == 'upbeat' and 'πΉ Instrumental' in use_cases:
score += 6
# Stem compatibility
model_stems = str(model_info).lower()
for stem in desired_stems:
if stem.lower() in model_stems:
score += 5
# Architecture preference based on priority
arch_type = model_info.get('architecture_type', '')
if priority == "quality" and "Roformer" in arch_type:
score += 15
elif priority == "speed" and "MDX-Net" in arch_type:
score += 15
model_scores[model_name] = score
# Return highest scoring model
if model_scores:
best_model = max(model_scores.items(), key=lambda x: x[1])
return best_model[0]
return None
except Exception as e:
self.logger.error(f"Error in auto-select: {str(e)}")
return None
def compare_models(self, audio_file: str, model_list: List[str]) -> Dict:
"""Enhanced model comparison with detailed metrics"""
if not audio_file or not model_list:
return {"error": "Please provide audio file and select models to compare"}
comparison_results = {
"audio_analysis": self.analyze_audio_characteristics(audio_file),
"model_results": {},
"summary": {},
"recommendations": []
}
for model_name in model_list:
try:
start_time = time.time()
# Initialize separator for this model
success, message = self.initialize_separator(model_name)
if not success:
comparison_results["model_results"][model_name] = {
"status": "Failed",
"error": message,
"processing_time": 0
}
continue
# Process audio
output_files = self.separator.separate(audio_file)
processing_time = time.time() - start_time
# Analyze results
if output_files and os.path.exists(output_files[0]):
audio_data, sample_rate = sf.read(output_files[0])
# Calculate quality metrics
quality_metrics = self._calculate_quality_metrics(audio_data, sample_rate)
comparison_results["model_results"][model_name] = {
"status": "Success",
"processing_time": processing_time,
"output_files": len(output_files),
"sample_rate": sample_rate,
"duration": len(audio_data) / sample_rate,
"quality_metrics": quality_metrics,
"output_stems": [os.path.basename(f) for f in output_files],
"model_info": self.get_available_models().get(model_name, {})
}
# Clean up
for file_path in output_files:
if os.path.exists(file_path):
os.remove(file_path)
else:
comparison_results["model_results"][model_name] = {
"status": "Failed",
"error": "No output files generated",
"processing_time": processing_time
}
except Exception as e:
comparison_results["model_results"][model_name] = {
"status": "Error",
"error": str(e),
"processing_time": 0
}
# Generate summary and recommendations
comparison_results["summary"] = self._generate_comparison_summary(comparison_results["model_results"])
comparison_results["recommendations"] = self._generate_recommendations(
comparison_results["audio_analysis"],
comparison_results["model_results"]
)
return comparison_results
def _calculate_quality_metrics(self, audio_data: np.ndarray, sample_rate: int) -> Dict:
"""Calculate audio quality metrics"""
try:
# RMS level
rms = np.sqrt(np.mean(audio_data**2))
# Dynamic range
peak = np.max(np.abs(audio_data))
dynamic_range = 20 * np.log10(peak / (rms + 1e-10))
# Spectral characteristics
spectral_centroid = np.mean(librosa.feature.spectral_centroid(y=audio_data, sr=sample_rate))
return {
"rms_level": float(rms),
"peak_level": float(peak),
"dynamic_range": float(dynamic_range),
"spectral_centroid": float(spectral_centroid),
"length_samples": len(audio_data),
"length_seconds": len(audio_data) / sample_rate
}
except Exception as e:
return {"error": str(e)}
def _generate_comparison_summary(self, model_results: Dict) -> Dict:
"""Generate summary statistics from model comparison"""
successful_results = {k: v for k, v in model_results.items() if v.get("status") == "Success"}
if not successful_results:
return {"message": "No successful model runs to compare"}
summary = {
"total_models": len(model_results),
"successful_models": len(successful_results),
"fastest_model": None,
"slowest_model": None,
"best_quality": None,
"average_processing_time": 0
}
# Find fastest and slowest
if successful_results:
times = {k: v.get("processing_time", 0) for k, v in successful_results.items()}
summary["fastest_model"] = min(times.items(), key=lambda x: x[1])[0]
summary["slowest_model"] = max(times.items(), key=lambda x: x[1])[0]
summary["average_processing_time"] = np.mean(list(times.values()))
return summary
def _generate_recommendations(self, audio_analysis: Dict, model_results: Dict) -> List[str]:
"""Generate intelligent recommendations based on comparison"""
recommendations = []
# Find best performing model
successful_models = {k: v for k, v in model_results.items() if v.get("status") == "Success"}
if successful_models:
# Find fastest successful model
fastest_model = min(successful_models.items(),
key=lambda x: x[1].get("processing_time", float('inf')))
recommendations.append(f"β‘ Fastest: {fastest_model[0]} ({fastest_model[1]['processing_time']:.2f}s)")
# Find model with most outputs
most_outputs = max(successful_models.items(),
key=lambda x: x[1].get("output_files", 0))
recommendations.append(f"ποΈ Most stems: {most_outputs[0]} ({most_outputs[1]['output_files']} files)")
# Audio-based recommendations
audio_type = audio_analysis.get('audio_type', 'unknown')
if audio_type == 'bass_heavy':
recommendations.append("πΈ Consider models with bass separation capabilities")
elif audio_type == 'bright_crisp':
recommendations.append("π€ Models optimized for vocal clarity work best")
elif audio_type == 'upbeat':
recommendations.append("πΉ Fast processing models recommended for energetic tracks")
return recommendations
def initialize_separator(self, model_name: str = None, **kwargs):
"""Initialize the separator with specified parameters"""
try:
with self.model_lock:
# Clean up previous separator if exists
if self.separator is not None:
del self.separator
torch.cuda.empty_cache()
# Set default model if not specified
if model_name is None:
models = self.get_available_models()
if models:
model_name = list(models.keys())[0] # Use first available model
else:
return False, "No models available"
# Initialize separator with updated parameters
self.separator = Separator(
output_format="WAV",
use_autocast=True,
use_soundfile=True,
**kwargs
)
# Load the model
self.separator.load_model(model_name)
self.current_model = model_name
return True, f"Successfully initialized with model: {model_name}"
except Exception as e:
self.logger.error(f"Error initializing separator: {str(e)}")
return False, f"Error initializing separator: {str(e)}"
def infer(self, audio_file: str, model_name: str, output_format: str = "WAV",
quality_preset: str = "Standard", custom_params: Dict = None,
enable_auto_optimize: bool = True):
"""Enhanced audio processing with auto-optimization"""
if audio_file is None:
return None, "No audio file provided"
if model_name is None:
return None, "No model selected"
# Auto-optimize parameters if enabled
if enable_auto_optimize:
audio_analysis = self.analyze_audio_characteristics(audio_file)
custom_params = self._optimize_parameters_for_audio(audio_analysis, custom_params)
if self.separator is None or self.current_model != model_name:
success, message = self.initialize_separator(model_name)
if not success:
return None, message
try:
start_time = time.time()
# Apply quality preset
if custom_params is None:
custom_params = {}
if quality_preset == "Fast":
custom_params.update({
"mdx_params": {"batch_size": 4, "overlap": 0.1, "segment_size": 128},
"vr_params": {"batch_size": 8, "aggression": 3},
"demucs_params": {"shifts": 1, "overlap": 0.1},
"mdxc_params": {"batch_size": 4, "overlap": 4}
})
elif quality_preset == "High Quality":
custom_params.update({
"mdx_params": {"batch_size": 1, "overlap": 0.5, "segment_size": 512, "enable_denoise": True},
"vr_params": {"batch_size": 1, "aggression": 8, "enable_tta": True, "enable_post_process": True},
"demucs_params": {"shifts": 4, "overlap": 0.5, "segments_enabled": False},
"mdxc_params": {"batch_size": 1, "overlap": 16, "pitch_shift": 0}
})
# Update separator parameters
for key, value in custom_params.items():
if hasattr(self.separator, key):
setattr(self.separator, key, value)
# Process the audio
output_files = self.separator.separate(audio_file)
processing_time = time.time() - start_time
# Read and prepare output audio
output_audio = {}
for file_path in output_files:
if os.path.exists(file_path):
# Create output with appropriate naming
stem_name = os.path.splitext(os.path.basename(file_path))[0]
audio_data, sample_rate = sf.read(file_path)
output_audio[stem_name] = (sample_rate, audio_data)
# Clean up file
os.remove(file_path)
if not output_audio:
return None, "No output files generated"
# Record processing history
history_entry = {
"timestamp": datetime.now().isoformat(),
"model": model_name,
"processing_time": processing_time,
"output_files": list(output_audio.keys()),
"audio_analysis": self.analyze_audio_characteristics(audio_file) if enable_auto_optimize else {},
"quality_preset": quality_preset
}
self.processing_history.append(history_entry)
return output_audio, f"Processing completed in {processing_time:.2f}s with model: {model_name}"
except Exception as e:
error_msg = f"Error processing audio: {str(e)}"
self.logger.error(f"{error_msg}\n{traceback.format_exc()}")
return None, error_msg
def _optimize_parameters_for_audio(self, audio_analysis: Dict, custom_params: Dict) -> Dict:
"""Automatically optimize parameters based on audio characteristics"""
if custom_params is None:
custom_params = {}
# Adjust parameters based on audio characteristics
duration = audio_analysis.get('duration', 0)
audio_type = audio_analysis.get('audio_type', 'balanced')
# For longer audio, increase batch size for efficiency
if duration > 300: # 5 minutes
custom_params.setdefault('mdx_params', {})['batch_size'] = 2
custom_params.setdefault('vr_params', {})['batch_size'] = 2
# For bass-heavy audio, increase aggression
if audio_type == 'bass_heavy':
custom_params.setdefault('vr_params', {})['aggression'] = 7
# For bright/crisp audio, enable post-processing
if audio_type == 'bright_crisp':
custom_params.setdefault('vr_params', {})['enable_post_process'] = True
# For dynamic audio, enable TTA for better quality
if audio_analysis.get('dynamic_range', 0) > 0.1:
custom_params.setdefault('vr_params', {})['enable_tta'] = True
return custom_params
def get_phistory(self):
"""Get enhanced processing history with analytics"""
if not self.processing_history:
return "No processing history available"
history_text = "π΅ Enhanced Processing History\n\n"
# Show recent entries with details
for i, entry in enumerate(self.processing_history[-10:], 1):
history_text += f"**{i}. {entry['timestamp'][:19]}**\n"
history_text += f" Model: {entry['model']}\n"
history_text += f" Time: {entry['processing_time']:.2f}s\n"
history_text += f" Stems: {', '.join(entry['output_files'])}\n"
# Add audio analysis if available
if 'audio_analysis' in entry and entry['audio_analysis']:
audio_type = entry['audio_analysis'].get('audio_type', 'unknown')
duration = entry['audio_analysis'].get('duration', 0)
history_text += f" Audio: {audio_type} ({duration:.1f}s)\n"
# Add quality preset info
if 'quality_preset' in entry:
history_text += f" Preset: {entry['quality_preset']}\n"
history_text += "\n"
return history_text
def reset_history(self):
"""Reset processing history"""
self.processing_history = []
return "Processing history cleared"
# Initialize the enhanced demo
demo1 = AudioSeparatorD()
# Create the Gradio interface directly
with gr.Blocks(theme="NeoPy/Soft", title="π΅ Enhanced Audio Separator") as app:
gr.Markdown(
"""
# π΅ Audio Separator Web UI
**Smart AI-Powered Audio Source Separation with Auto-Selection & Advanced Model Comparison**
β¨ **Features**: Auto model selection, performance analytics, smart parameter optimization, and comprehensive model comparison
"""
)
# System Information
with gr.Accordion("π₯οΈ System Information", open=False):
system_info = demo1.get_system_info()
info_text = f"""
**PyTorch Version:** {system_info['pytorch_version']}
**Hardware Acceleration:** {system_info['device'].upper()}
**CUDA Available:** {system_info['cuda_available']} (Version: {system_info['cuda_version']})
**Apple Silicon (MPS):** {system_info['mps_available']}
**GPU Memory:** {system_info['memory_allocated'] // 1024**2}MB / {system_info['memory_total'] // 1024**2}MB
"""
gr.Markdown(info_text)
with gr.Row():
with gr.Column():
# Main audio input
audio_input = gr.Audio(
label="π΅ Upload Audio File",
type="filepath"
)
# Add info text separately
gr.Markdown("*Upload audio for intelligent analysis and separation*")
# Auto-analyze button
analyze_btn = gr.Button("π Analyze Audio", variant="secondary")
# Audio analysis output
audio_analysis_output = gr.JSON(label="Audio Analysis Results", visible=False)
# Enhanced model selection
model_list = demo1.get_available_models()
# Model dropdown with enhanced display
model_dropdown = gr.Dropdown(
choices=list(model_list.keys()) if model_list else [],
value=list(model_list.keys())[0] if model_list else None,
label="π€ AI Model Selection",
elem_id="model_dropdown"
)
# Add info text separately
gr.Markdown("*Choose an AI model or use auto-selection*")
# Auto-selection controls
with gr.Row():
auto_select_btn = gr.Button("π― Auto-Select Best Model", variant="primary")
priority_radio = gr.Radio(
choices=["Quality", "Speed", "Balanced"],
value="Quality",
label="Selection Priority"
)
# Add info text separately
gr.Markdown("*What matters most for model selection?*")
# Model info display
model_info_display = gr.JSON(label="π Selected Model Information")
# Quality preset and optimization
with gr.Row():
quality_preset = gr.Radio(
choices=["Fast", "Standard", "High Quality", "Custom"],
value="Standard",
label="β‘ Processing Quality"
)
auto_optimize = gr.Checkbox(
label="π§ Auto-Optimize Parameters",
value=True
)
# Add info text separately
gr.Markdown("*Automatically optimize parameters based on audio analysis*")
# Enhanced advanced parameters
with gr.Accordion("π§ Advanced Parameters", open=False):
with gr.Row():
batch_size = gr.Slider(1, 8, value=1, step=1, label="Batch Size")
segment_size = gr.Slider(64, 1024, value=256, step=64, label="Segment Size")
overlap = gr.Slider(0.1, 0.5, value=0.25, step=0.05, label="Overlap")
with gr.Row():
denoise = gr.Checkbox(label="Enable Denoise", value=False)
tta = gr.Checkbox(label="Enable TTA", value=False)
post_process = gr.Checkbox(label="Enable Post-Processing", value=False)
pitch_shift = gr.Slider(-12, 12, value=0, step=1, label="Pitch Shift (semitones)")
# Process button
process_btn = gr.Button("π΅ Smart Separate Audio", variant="primary", size="lg")
with gr.Column():
# Status and results
status_output = gr.Textbox(label="π Status", lines=4)
# Enhanced output tabs
with gr.Tabs():
with gr.Tab("π€ Vocals"):
vocals_output = gr.Audio(label="Vocals")
with gr.Tab("πΉ Instrumental"):
instrumental_output = gr.Audio(label="Instrumental")
with gr.Tab("π₯ Drums"):
drums_output = gr.Audio(label="Drums")
with gr.Tab("πΈ Bass"):
bass_output = gr.Audio(label="Bass")
with gr.Tab("ποΈ Other Stems"):
other_output = gr.Audio(label="Other Stems")
# Performance metrics
performance_metrics = gr.JSON(label="π Performance Metrics", visible=False)
# Download section
with gr.Accordion("π₯ Batch & Download", open=False):
gr.Markdown("### π Batch Processing")
batch_files = gr.File(
file_count="multiple",
file_types=[".wav", ".mp3", ".flac", ".m4a"],
label="Batch Audio Files"
)
with gr.Row():
batch_btn = gr.Button("β‘ Process Batch")
auto_batch_btn = gr.Button("π― Auto-Select & Batch")
batch_output = gr.File(label="π¦ Download Batch Results")
# Enhanced Model Management Tabs
with gr.Tabs():
with gr.Tab("π Model Explorer"):
gr.Markdown("## π§ Intelligent Model Comparison & Selection")
# Enhanced model information
model_info = gr.JSON(value=demo1.get_available_models(), label="π Model Database")
refresh_models_btn = gr.Button("π Refresh Models")
# Advanced model filtering
with gr.Row():
filter_architecture = gr.Dropdown(
choices=["All", "MDX-Net", "Demucs", "Roformer", "VR Arch"],
value="All",
label="Filter by Architecture"
)
filter_use_case = gr.Dropdown(
choices=["All", "Vocals", "Instrumental", "Drums", "Bass", "Multi-stem"],
value="All",
label="Filter by Use Case"
)
filter_priority = gr.Dropdown(
choices=["All", "Quality", "Speed", "Memory Efficient"],
value="All",
label="Filter by Priority"
)
filtered_models = gr.Dropdown(
choices=list(model_list.keys())[:10] if model_list else [],
multiselect=True,
label="π― Models for Comparison"
)
# Add info text separately
gr.Markdown("*Select up to 5 models for detailed comparison*")
compare_btn = gr.Button("π¬ Advanced Model Comparison")
comparison_results = gr.JSON(label="π Comparison Results")
with gr.Tab("π Analytics & History"):
history_output = gr.Textbox(label="π Processing History", lines=15)
with gr.Row():
refresh_history_btn = gr.Button("π Refresh History")
reset_history_btn = gr.Button("ποΈ Clear History", variant="stop")
export_history_btn = gr.Button("π Export Analytics")
analytics_output = gr.JSON(label="π Analytics Dashboard")
with gr.Tab("π― Smart Recommendations"):
gr.Markdown("## π€ AI-Powered Model Recommendations")
recommendation_status = gr.Textbox(label="Recommendation Status", lines=3)
with gr.Row():
get_recommendations_btn = gr.Button("π― Get Smart Recommendations")
apply_recommendation_btn = gr.Button("β¨ Apply Best Recommendation")
recommendations_display = gr.JSON(label="π― Personalized Recommendations")
# Event handlers
def analyze_audio(audio_file):
if not audio_file:
return None, "No audio file provided"
analysis = demo1.analyze_audio_characteristics(audio_file)
# Format analysis for display
if "error" not in analysis:
formatted_analysis = f"""
**Audio Type:** {analysis.get('audio_type', 'Unknown').title().replace('_', ' ')}
**Duration:** {analysis.get('duration', 0):.1f} seconds
**Sample Rate:** {analysis.get('sample_rate', 0)} Hz
**Tempo:** {analysis.get('tempo', 0):.1f} BPM
**Spectral Characteristics:** {analysis.get('avg_spectral_centroid', 0):.0f} Hz (centroid)
**Dynamic Range:** {analysis.get('dynamic_range', 0):.3f}
"""
return analysis, formatted_analysis
else:
return analysis, f"Analysis failed: {analysis['error']}"
def auto_select_model(audio_file, priority):
if not audio_file:
return None, "No audio file provided", None
# Analyze audio first
audio_analysis = demo1.analyze_audio_characteristics(audio_file)
# Determine desired stems based on audio analysis
desired_stems = ["vocals"] # Default
if audio_analysis.get('audio_type') == 'bass_heavy':
desired_stems.append("bass")
elif audio_analysis.get('tempo', 0) > 120:
desired_stems.append("drums")
# Auto-select model
selected_model = demo1.auto_select_model(
audio_analysis, desired_stems, priority.lower()
)
if selected_model:
models = demo1.get_available_models()
model_info = models.get(selected_model, {})
return (
selected_model,
f"π― Auto-selected: {model_info.get('friendly_name', selected_model)}\n"
f"Architecture: {model_info.get('architecture_type', 'Unknown')}\n"
f"Best for: {', '.join(model_info.get('use_cases', [])[:2])}",
model_info
)
else:
return None, "Auto-selection failed - no suitable model found", None
def update_model_info(model_name):
if not model_name:
return None
models = demo1.get_available_models()
model_info = models.get(model_name, {})
if model_info:
# Format model information
friendly_info = {
"π€ Friendly Name": model_info.get('friendly_name', model_name),
"ποΈ Architecture": model_info.get('architecture_type', 'Unknown'),
"π‘ Best For": model_info.get('use_cases', []),
"β‘ Performance": model_info.get('performance_characteristics', {}),
"π― Recommendations": model_info.get('recommended_for', {}),
"π Technical Details": {
"Filename": model_name,
"Supported Stems": len(str(model_info)) // 10 # Rough estimate
}
}
return friendly_info
return {"error": "Model information not available"}
def infer(audio_file, model_name, quality_preset, batch_size, segment_size,
overlap, denoise, tta, post_process, pitch_shift, auto_optimize):
if not audio_file or not model_name:
return None, None, None, None, None, "Please upload an audio file and select a model", None
# Prepare custom parameters
custom_params = {
"mdx_params": {
"batch_size": int(batch_size),
"segment_size": int(segment_size),
"overlap": float(overlap),
"enable_denoise": denoise
},
"vr_params": {
"batch_size": int(batch_size),
"enable_tta": tta,
"enable_post_process": post_process,
"aggression": 5 # Default
},
"demucs_params": {
"overlap": float(overlap)
},
"mdxc_params": {
"batch_size": int(batch_size),
"overlap": int(overlap * 10),
"pitch_shift": int(pitch_shift)
}
}
output_audio, status = demo1.infer(
audio_file, model_name,
quality_preset=quality_preset,
custom_params=custom_params,
enable_auto_optimize=auto_optimize
)
if output_audio is None:
return None, None, None, None, None, status, None
# Extract different stems
vocals = None
instrumental = None
drums = None
bass = None
other = None
for stem_name, (sample_rate, audio_data) in output_audio.items():
if "vocal" in stem_name.lower():
vocals = (sample_rate, audio_data)
elif "instrumental" in stem_name.lower():
instrumental = (sample_rate, audio_data)
elif "drum" in stem_name.lower():
drums = (sample_rate, audio_data)
elif "bass" in stem_name.lower():
bass = (sample_rate, audio_data)
else:
other = (sample_rate, audio_data)
# Generate performance metrics
performance_metrics = {
"Model": model_name,
"Quality Preset": quality_preset,
"Output Stems": len(output_audio),
"Processing": "Completed Successfully"
}
return vocals, instrumental, drums, bass, other, status, performance_metrics
def compare_models_advanced(audio_file, model_list):
if not audio_file or not model_list:
return {"error": "Please upload an audio file and select models to compare"}
results = demo1.compare_models(audio_file, model_list)
return results
def get_smart_recommendations(audio_file):
if not audio_file:
return "Please upload an audio file first", {}
# Analyze audio
audio_analysis = demo1.analyze_audio_characteristics(audio_file)
models = demo1.get_available_models()
# Generate recommendations
recommendations = {
"audio_analysis": audio_analysis,
"recommended_models": [],
"tips": []
}
# Quality-focused recommendations
quality_models = []
speed_models = []
for model_name, model_info in models.items():
perf_chars = model_info.get('performance_characteristics', {})
if perf_chars.get('quality_rating') == 'high':
quality_models.append({
'model': model_name,
'name': model_info.get('friendly_name', model_name),
'reason': 'High quality output'
})
if perf_chars.get('speed_rating') == 'fast':
speed_models.append({
'model': model_name,
'name': model_info.get('friendly_name', model_name),
'reason': 'Fast processing'
})
recommendations["recommended_models"] = {
"π― For Best Quality": quality_models[:3],
"β‘ For Speed": speed_models[:3]
}
# Audio-specific tips
audio_type = audio_analysis.get('audio_type', 'balanced')
if audio_type == 'bass_heavy':
recommendations["tips"].append("πΈ Models with bass separation work best")
elif audio_type == 'bright_crisp':
recommendations["tips"].append("π€ Post-processing enabled for vocal clarity")
elif audio_type == 'upbeat':
recommendations["tips"].append("π₯ Consider drum isolation for energetic tracks")
status = f"β
Generated recommendations for {audio_analysis.get('audio_type', 'unknown')} audio"
return status, recommendations
def apply_best_recommendation(audio_file):
if not audio_file:
return None, "Please upload an audio file first", None
# Get auto-selection with quality priority
audio_analysis = demo1.analyze_audio_characteristics(audio_file)
selected_model = demo1.auto_select_model(
audio_analysis, ["vocals"], "quality"
)
if selected_model:
models = demo1.get_available_models()
model_info = models.get(selected_model, {})
return (
selected_model,
f"β¨ Applied recommendation: {model_info.get('friendly_name', selected_model)}",
model_info
)
else:
return None, "Could not generate recommendations", None
# Wire up event handlers
analyze_btn.click(
fn=analyze_audio,
inputs=[audio_input],
outputs=[audio_analysis_output, recommendation_status]
)
auto_select_btn.click(
fn=auto_select_model,
inputs=[audio_input, priority_radio],
outputs=[model_dropdown, recommendation_status, model_info_display]
)
model_dropdown.change(
fn=update_model_info,
inputs=[model_dropdown],
outputs=[model_info_display]
)
process_btn.click(
fn=infer,
inputs=[
audio_input, model_dropdown, quality_preset,
batch_size, segment_size, overlap, denoise, tta, post_process,
pitch_shift, auto_optimize
],
outputs=[
vocals_output, instrumental_output, drums_output,
bass_output, other_output, status_output, performance_metrics
]
)
compare_btn.click(
fn=compare_models_advanced,
inputs=[audio_input, filtered_models],
outputs=[comparison_results]
)
refresh_models_btn.click(
fn=lambda: demo1.get_available_models(),
outputs=[model_info]
)
refresh_history_btn.click(
fn=lambda: demo1.get_phistory(),
outputs=[history_output]
)
reset_history_btn.click(
fn=lambda: demo1.reset_history(),
outputs=[history_output]
)
get_recommendations_btn.click(
fn=get_smart_recommendations,
inputs=[audio_input],
outputs=[recommendation_status, recommendations_display]
)
apply_recommendation_btn.click(
fn=apply_best_recommendation,
inputs=[audio_input],
outputs=[model_dropdown, recommendation_status, model_info_display]
)
# Batch processing
def batch_inf(batch_files, model_name):
if not batch_files or not model_name:
return None, "Please upload batch files and select a model"
import zipfile
import io
zip_buffer = io.BytesIO()
with zipfile.ZipFile(zip_buffer, 'w', zipfile.ZIP_DEFLATED) as zip_file:
for file_info in batch_files:
output_audio, _ = demo1.infer(file_info, model_name)
if output_audio is not None:
for stem_name, (sample_rate, audio_data) in output_audio.items():
import tempfile
with tempfile.NamedTemporaryFile(suffix='.wav', delete=False) as tmp_file:
sf.write(tmp_file.name, audio_data, sample_rate)
with open(tmp_file.name, 'rb') as f:
zip_file.writestr(f"{os.path.splitext(os.path.basename(file_info))[0]}_{stem_name}.wav", f.read())
os.unlink(tmp_file.name)
zip_buffer.seek(0)
return gr.File(value=zip_buffer, visible=True), f"Batch processing completed for {len(batch_files)} files"
batch_btn.click(
fn=batch_inf,
inputs=[batch_files, model_dropdown],
outputs=[batch_output, status_output]
)
def auto_batch_process(batch_files, priority):
if not batch_files:
return None, "Please upload batch files"
# Auto-select best model for first file as representative
if batch_files:
audio_analysis = demo1.analyze_audio_characteristics(batch_files[0])
selected_model = demo1.auto_select_model(audio_analysis, ["vocals"], priority.lower())
if selected_model:
return batch_inf(batch_files, selected_model)
return None, "Auto-selection failed"
auto_batch_btn.click(
fn=auto_batch_process,
inputs=[batch_files, priority_radio],
outputs=[batch_output, status_output]
)
app.launch(
server_port=7860,
share=True,
ssr_mode=True
) |